Skill Claude Code needs its repo
Carry out one planned task and review the result. Use this skill when tasks have been detailed and you are ready for the work to actually be done. It runs the task according to its type -- RESEARCH, DESIGN, IMPLEMENT, TEST, DOCUMENT, MIGRATE, REFACTOR or CONFIGURE -- then loops through review and fixes until the…
Skill Claude Code
Design, run, and analyze a controlled experiment or A/B test. Use this skill when you already know what to experiment and want it scoped, specified, and broken into reviewable tasks in a single pass, with no requirements interview. You approve the resulting plan before anything is built: this skill plans and stops, so…
Skill Claude Code
Diagnose and correct a defect, regression, incident, or vulnerability. Use this skill when you already know what to fix and want it scoped, specified, and broken into reviewable tasks in a single pass, with no requirements interview. You approve the resulting plan before anything is built: this skill plans and stops…
Skill Claude Code needs its repo
Sweep the project back into a consistent state after work has landed. Use this skill when the Knowledge Base has drifted from the code, the generated summary is stale, or work-area artifacts have piled up. It runs three gated jobs in order: re-discovering KB docs that changed since the last approval, regenerating the…
Skill Claude Code
Migrate data, a dependency, framework, or platform, with a rollback plan (non-schema; schema migrations use create/update-data-model). Use this skill when you already know what to migrate and want it scoped, specified, and broken into reviewable tasks in a single pass, with no requirements interview. You approve the…
Skill Claude Code
Watch production, classify what you find, and route it to whoever should act. Use this skill after a deployment, on a schedule, or whenever something looks wrong in the field. It reads telemetry, detects anomalies, performs root-cause analysis on the ones that are bugs, and then routes: bugs to /aid-fix, change…
Skill Claude Code
Sequence feature SPECs into deliverables -- each one a functional MVP that builds on the previous. Strategy, not tactics. Use when feature SPECs are complete and you need a delivery roadmap.
Skill Claude Code
Build a throwaway, low-fidelity UI wireframe and interaction flow to test whether a UX direction actually works. Use this skill when you want to see a screen before committing to building it. Isolated and disposable: it never touches production, and it reports what the mock shows rather than deciding anything for you.…
Skill Claude Code
Build a throwaway, low-fidelity model to test whether a direction actually works, before anyone commits to building it properly. Use this skill when a decision hinges on something you can only learn by trying it. The model is isolated and disposable -- it lives in the work folder or an opt-in worktree and never…
Skill Claude Code
Read one ticket from the project's issue tracker and show its fields. Use this skill when you need a ticket's current contents -- during triage, before starting work, or while writing a change that references it. Its grammar is aid-read-ticket [ :] -- the tracker's own id, optionally prefixed with a connector name…
Skill Claude Code
Restructure or optimize code without changing behavior (rename, restructure, or improve performance). Use this skill when you already know what to refactor and want it scoped, specified, and broken into reviewable tasks in a single pass, with no requirements interview. You approve the resulting plan before anything is…
Skill Claude Code
Remove or delete a code artifact, endpoint, dependency, feature, or dead code; update dependents, tests, and docs. Use this skill when you already know what to remove and want it scoped, specified, and broken into reviewable tasks in a single pass, with no requirements interview. You approve the resulting plan before…
Skill Claude Code
Analyse data or usage and return a verified insight report in one pass -- exploratory analysis, metrics, or an A/B result. Use this skill when you have data and need to know what it says, with the caveats stated rather than buried. It presents findings, conclusions both positive and negative, data-quality caveats, any…
Skill Claude Code
Investigate an open technical question and return a verified answer in one pass -- evaluating options, or running an isolated feasibility spike if you explicitly authorise one. Use this skill when a decision is blocked on something nobody has established yet. It presents the answer, the conclusions both positive and…
Skill Claude Code
Review an existing artifact against criteria and return findings and recommendations in one pass -- code, a diff, a design, a pull request, a ticket, a document, a UI, whatever you name. Use this skill when you want a second opinion before something lands. It is read-only apart from its findings ledger: it never plans…
Skill Claude Code needs its repo
Add or update one entry in the connector catalog. Use this skill when a project gains a new external tool, or an existing connector's configuration changes, and you do not want to re-run discovery for it. Naming a tool and a type creates its descriptor when absent, or updates that same descriptor in place when…
Skill Claude Code
Turn one feature into a technical specification, collaboratively. Use this skill when a feature has been defined and how it will actually be built needs settling before any task is planned. The agent works as a tech lead: it reads the Knowledge Base, the requirements, and the codebase, proposes a technical approach…
Skill Claude Code
Generate kb.html, a single-file visual tour of the Knowledge Base. Use this skill when someone non-technical needs to read what the project knows about itself -- a newcomer, a stakeholder, anyone who will not open .aid/knowledge/ directly. It builds one section per resolved document from that document's own…
Skill Claude Code
Run data-quality checks against thresholds and report -- schema, freshness, completeness, uniqueness. Use this skill when you need to know whether a dataset or pipeline is currently trustworthy. Read-only; it resolves nothing, and findings hand off to /aid-fix. A thin kind-sibling of /aid-test, which defines its full…
Skill Claude Code
Run a performance verification against a threshold or SLO -- a benchmark, a load test, or a stress test -- and report measured against target. Use this skill when you need to know whether something is fast enough, and by how much. Read-only; it resolves nothing, and findings hand off to /aid-fix. A thin kind-sibling…
Skill Claude Code
Run a security verification and consolidate the findings -- SAST, DAST, fuzzing, or a dependency audit. Use this skill when you need to know what is currently exploitable or outdated. Read-only; it resolves nothing, and findings hand off to /aid-fix. A thin kind-sibling of /aid-test, which defines its full behavior.
Skill Claude Code
Run a test suite or verification and consolidate the results into findings, in one pass. Use this skill when you need to know the current state of something measurable -- unit, integration or end-to-end tests, a security scan, a performance benchmark, a data-quality check, or a model evaluation. It runs whatever the…
Skill Claude Code
Suggest which AID entry point fits the work you are describing. Use this skill when you know what you want to change but not which skill to reach for. Give it one short free-form description; it infers the work type, judges the scope, and names the single best entry -- the matching verb-and-artifact shortcut for a…
Skill Claude Code needs its repo
Remove one entry from the connector catalog. Use this skill when a project stops using an external tool and its descriptor and stored secret should go with it. Naming the tool deletes its descriptor, purges its secret, and rebuilds the catalog index from whatever remains on disk -- touching only that one stem, so…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: